EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0881
Title: Cryptocurrencies' quantile and tail expectation forecasting Authors:  Kokulo Kpai Lawuobahsumo - University of Calabria (Italy) [presenting]
Bernardina Algieri - University of Calabria (Italy)
Arturo Leccadito - Università della Calabria (Italy)
Abstract: The aim is to jointly predict conditional quantiles and tail expectations for the returns of the most popular cryptocurrencies (Bitcoin, Dogecoin, Ethereum, Ripple, and Litecoin) using financial and macroeconomic indicators as explanatory variables. The financial variables we use are Nasdaq Composite, WTI Futures, Gold Fixing Price 3:00 P.M. (London time), and CBOE Volatility Index (VIX). The economic variables we considered are U.S. Dollar Index, 5-Year Forward Inflation Expectation Rate, 10-Year Breakeven Inflation Rate, 10-Year Treasury Constant Maturity, and 10-Year Market Yield on U.S. Treasury Securities. We use daily data and the Monotone Composite quantile regression neural network model (MCQRNN) to make one-step-ahead and five-step-ahead predictions for Value-at-Risk (VaR) and Expected Shortfall (ES) on a rolling basis and compare the performance of our model against the standard GARCH (1,1) model. The superior set of models is then chosen by backtesting $\alpha$-VaR and $\alpha$-ES using a Model Confidence Set (MCS) procedure with a loss function. Our results show that the MCQRNN performed better than the benchmark model for jointly predicting VaR and ES. The result is consistent for 1\% and 5\% levels of $\alpha$ both in the right and left tails for all cryptocurrencies.